Current state of inbound fees

[DerEwige]

📌 Current State of Inbound Fees

Hello @t-bast,

In the meeting yesterday I mentioned I’d make a short post regarding my view about the current BLIP18 inbound fee situation.

I’ll introduce the concept of “fee potential” to make things easier to explain. The section below covers it — if you’re already familiar, feel free to skip ahead.

---

🔎 Introduction to Fee Potential

In Lightning, we often use the terms “source” and “sink”:

• A source provides liquidity.
• A sink consumes liquidity.
• These terms are relative, meaning the same node can be a source to one node and a sink to another.

The fee potential (FP) attempts to quantify this relation into a single number:
👉 It is the expected fee (ppm) required to reach node B from node A.

Notation:
FP(A -> B)

Notes:

• A node always has FP(A -> A) = 0.
• A positive FP means that B is a sink relative to A.
• A negative FP means that B is a source relative to A.

📊 Fee Potential Table

This is a list of fee potentials as measured from my node for a few selected nodes.

Node	Fee Potential
LOOP	3486
bfx-lnd0	3055
Kraken 🐙⚡	2280
ACINQ	1901
LNBiG [Hub-1]	1867
Strike	1800
WalletOfSatoshi.com	1422
Boltz	1381
cyberdyne.sh	613
carlslight	271
speedupln.com	0

📊 scatter plot

Examples

• FP(speedupln.com -> Kraken 🐙⚡) = 2280
  → Kraken is a strong sink relative to my node.

• FP(B -> A) = - FP(A -> B)
  So:
  FP(Kraken 🐙⚡ -> speedupln.com) = -2280
  → My node is a strong source to Kraken.

• General formula:
  FP(B -> C) = FP(A -> C) - FP(A -> B)

  Example:
  FP(ACINQ -> bfx-lnd0) = FP(speedupln.com -> bfx-lnd0) - FP(speedupln.com -> ACINQ)
= 3055 - 1901 = 1154

  → bfx-lnd0 is a sink for ACINQ.

⚠️ Note: the lowest real fee in the network is 0 ppm. Negative fee potential values collapse to ~0 ppm in practice.

---

📉 Inbound Fee Situation

• Most natural traffic flows from a source → to a sink.
• This means payments flow from lower FP → to higher FP.
• If senders cannot use negative inbound fees, they will not route through nodes with strongly negative FP, because this would require higher fees for the total route and not be economical.
• Thus, “deep negative FP” nodes effectively become invisible to senders.

Current numbers

• 220 nodes currently offer negative inbound fees (≈ 4000 channels).
• 8 nodes demand positive inbound fees (≈ 40 channels).

Most of these nodes are below ACINQ in FP — making them unlikely to be routed through.

---

📈 Positive Inbound Fees

• The most notable example is cyberdyne.sh, which only charges inbound fees on peers below it in FP (who don’t offer negative inbound fees back).
• This makes it even less likely for ACINQ to route through such channels.

So far:

• Positive inbound fees are confined to low fee‑potential regions.
• They currently have little network impact.

⚠️ But: if a large high‑FP node (e.g., LNBiG) adopted them, the effect would be major.

---

📉 Negative Inbound Fees

• Many small–medium routing nodes use negative inbound fees to stay competitive.

Example route:
ACINQ → bfx-lnd0 → speedupln.com → Loop

• Cost: ~1500 ppm
• Much cheaper than the direct 3000 ppm charged by ACINQ
• Close to predicted FP, with ~100 ppm margin

Larger adopters:

• Strike
• cyberdyne.sh
• Many other smaller nodes at lower FP levels

Why?

Negative inbound fees allow nodes to:

• Adjust fees based on FP differences between peers
• Enable natural flow through nearly all peers
• (Except: typically no inbound flow from the strongest sink)

---

✅ TL;DR

• Positive inbound fees: currently isolated to low‑FP regions, no major impact yet.

  • ⚠️ Risk: if a large high‑FP node adopts them, major disruptions are possible.

• Negative inbound fees: thriving in low‑FP regions.

  • 📉 High‑FP nodes could greatly benefit from adopting them (lower routing costs, more natural flow).

---

🗣️ A Few Last Words

I believe the risk of a high‑FP positive inbound node is real.
Implementations without BLIP18 support would be caught off guard by this situation.

I think the different implementations should have a discussion about this.
For example, LND could apply guardrails by excluding positive inbound fee channels from route finding while accept-positive-inbound-fees is set to false.
This would instantly kill all attempts to introduce positive inbound fees as no sender (even LND in default settings) could use them by default.